On Jan 23, 2008 12:10 PM, Thomas Bächler <[EMAIL PROTECTED]> wrote: > It is a much requested feature: Someone wants to search for a filename > and get a package name, and there is no possibility in Arch. > > I have scripts to generate filelists from the package files which could > run once a day or so. However, cactus (after a short jabber discussion) > disagress with me on how the search would be done: > > My suggestion: > Create a script on the webserver that "greps" through the prepared > filelists or searches a sql database, then gives you a machine-readable > output via xml or json that can be displayed in a client (for example a > 'pacman -So' option). > However, cactus thinks that this would put too much load on the server. > > Another possibility: Make the filelists available for (optional) > download and search them offline (with pacman or another tool). My > problem here is: > total 140M > -rw-r--r-- 1 thomas users 35M Jan 21 17:06 filelist.community-i686 > -rw-r--r-- 1 thomas users 34M Jan 21 17:11 filelist.community-x86_64 > -rw-r--r-- 1 thomas users 1.6M Jan 21 16:52 filelist.core-i686 > -rw-r--r-- 1 thomas users 1.6M Jan 21 16:52 filelist.core-x86_64 > -rw-r--r-- 1 thomas users 32M Jan 21 16:57 filelist.extra-i686 > -rw-r--r-- 1 thomas users 32M Jan 21 17:02 filelist.extra-x86_64 > -rw-r--r-- 1 thomas users 2.1M Jan 21 17:11 filelist.testing-i686 > -rw-r--r-- 1 thomas users 2.0M Jan 21 17:12 filelist.testing-x86_64 > -rw-r--r-- 1 thomas users 1.2M Jan 21 17:11 filelist.unstable-i686 > -rw-r--r-- 1 thomas users 673K Jan 21 17:11 filelist.unstable-x86_64 > > Which is 8.9MB compressed with bz2, a lot to download if you have to > update the lists every day to have an up-to-date version. > > What do you guys think?
I would implement this as follows: Add some option to repo-add to generate a second list like above. "repo-add --filelists" would generate extra.files.tar.gz or something Then a pacman config option like so: SyncFilelists = yes And everything else should flow from there. Integrating json and xml parsing into pacman just seems.... wrong